transformer build thread

[abraxalito]

You had two trafos with you - were both designed for smartphone use?

Yes all the transformers I brought with me from China were designed for smartphones. I hadn't imagined trafos would be such a hit in a hi-fi system too :)

Yea, well I was thinking that the winding costs of such very thin wire my be significantly different (both in number of turns & breakages) compared to thicker wire?

Could be, more turns takes more time that's for sure. But I'd be concerned about releasing to the market a trafo which would measure extremely poor in a frequency response plot at full level should someone decide to make one. Or that overloaded on some particularly bass-heavy music.

There's one more factor relating to transformer isolation that occurred to me after the last post - the conductivity of the core. High permeability types (10K for example) tend to be more conductive so it maybe that PC40 provides the best isolation. Its another factor to look into. Parasitic capacitances I've not seen mentioned on any steel/nickel cored transformer manufacturer's datasheet but our little experiment demonstrated they're important. Lower capacitance is going to reject CM noise better and when I made a measurement into interwinding capacitance on a trafo I was taking apart I found it was dominated by winding-core capacitance.

[abraxalito]

Primary windings for the second prototype I managed to complete without any breakage. The DCRs are roughly 720 and 800ohms - the latter one is on the outside and hence needs more wire. I'm also using the inner pins on the bobbin only as the wires to the outer pins are much more vulnerable to breakage when the core is fitted.

Of interest to me is the parasitic shunt capacitance of each winding which is best measured before any core is fitted to the bobbin. I get a rough estimate of this by looking on my LCR meter how the reactance varies with frequency. In this case with the LCR meter set to measure inductance, it changes to negative between 50k and 100kHz. A negative inductance means the shunt capacitance dominates over the inductance. So somewhere between 50 and 100k we have the inductance resonating with the shunt capacitance, this leads to a first stab of 150pF for the shunt capacitance. There is going to be some mutual inductance between the windings even with no core fitted so it doesn't make too much sense to try for a much more accurate figure than this guided guess.

I'll add a couple of layers of yellow transformer tape next to space the windings apart (aiming to lower the inter-winding capacitance) then spin on the secondary...

[abraxalito]

Twin secondaries went on without issue, I measured the DCRs at 180ohm and 200ohm. Armed with all the DCRs I'll later do a check to see if I've chosen the optimum wire diameters.

I've fitted the two halves of the core, some jiggling of them is in order to get them to mate together with the maximum contact area, this then gives the highest possible inductance. Cleaning the surfaces can help, or at least wiping them with a microfibre cloth to ensure as few contaminants (oil from fingers for example) are left on the surface. Then I strap the two halves up with a couple of loops of transformer tape. Voila.

Primary inductance at 50Hz, 1V excitation shows as 17H. This shouldn't be at all hard to drive, the inductance is in parallel with around 150pF, so not too much different from a length of normal audio IC.

[jkeny]

Great work - nice progress report & comments - I'll have to mull over all you say - I'm not fluent in trafo speak yet :)

As Nige says, the finished product does look beautiful.

These are quite a bit bigger than the smartphone trafos - about 4 times the size or more?

[abraxalito]

The trafos you got inside the ally box I think were EE19 cores, not a huge difference, maybe 50% more volume? You can look up the dimensions if you download the Ferroxcube databook, PQ20/20 cores are on page 722.

I'm running a test to see how much interwinding capacitance there is and I'd like to compare with the ones we listened to at Nige's.

[jkeny]

The trafos you got inside the ally box I think were EE19 cores, not a huge difference, maybe 50% more volume? You can look up the dimensions if you download the Ferroxcube databook, PQ20/20 cores are on page 722.

Oh, OK I was judging it by what I thought was a CD that it was sitting on but I see now that it's a reel of wire - I can only blame my cold for my extra dose of stupidity :)

I'm running a test to see how much interwinding capacitance there is and I'd like to compare with the ones we listened to at Nige's.

Great - look forward to those results

I hadn't imagined trafos would be such a hit in a hi-fi system too :)

Well, on TnHF you won't get the stock response about trafos that is often found elsewhere "Well it must be distortion that's liked, blah, blah" - we are more pragmatic in our approach & if it sounds "right" then it is right, even if we don't fully understand why.

[abraxalito]

I measured the new trafo and the interwinding capacitance is in the region 50 - 100pF. Which is fine but I am curious to see if I can get it lower and an idea occurred to me that perhaps a signal common-mode choke might help. The CM choke being designed to give isolation at high frequencies (where the transformer is a bit poor) and not try to isolate at all (where the transformer is excellent) at low frequencies.

Here's my second attempt, the first attempt didn't get far as the capacitance was lower than the trafo but not very much lower (just under 50pF). A few years ago I made some great power line CM chokes using these segmented bobbins so I figured it was worth a try.

[jkeny]

OK, just trying to follow along here - lower capacitance = better common mode noise rejection = lower/no isolation @LF but good isolation @ HF?
We premise that the SQ benefits heard in Nige's were mainly due CM noise rejection rather than other noise rejection. This suggests HF CM noise was the culprit, right?

Have you got a playback setup that emulates the SQ benefits heard in Nige's with the original trafo's (that you still have)?
This would enable some real world testing of the premise

The CM chokes look far easier to wind being multi-segmented?

[abraxalito]

OK, just trying to follow along here - lower capacitance = better common mode noise rejection = no isolation.
We premise that the SQ benefits heard in Nige's were mainly due CM noise rejection rather than other noise rejection.

I'm not up to speed on what 'other noise' might be here. Yes seems Nige's amp was susceptible to the CM noise from the computer/monitor which was strongly blocked by the trafos.

Have you got a playback setup that emulates the SQ benefits heard in Nige's with the original trafo's (that you still have)?
This would enable some real world testing of the premise

I have an amp which doesn't already have trafos inside. But all my DACs have had trafos on the output from at least 3 years ago, so I suppose I could try with a smartphone or even with an el-cheapo SD-card player using a noisy PSU.

The CM chokes look far easier to wind being multi-segmented?

Easier than the trafos? Pretty similar, though here the wire's 0.1mm and could be even thicker.

The plan is to put the CM choke in series with the trafo to extend the isolation top end by a few octaves. I think I can reduce the trafo's capacitance by about an order of magnitude by means of the CM choke, so down to 10pF max. What I really need now is a supply of pre-twisted-pair wire so I can try out a bifilar CM choke....

[jkeny]

OK, just trying to follow along here - lower capacitance = better common mode noise rejection = no isolation.
We premise that the SQ benefits heard in Nige's were mainly due CM noise rejection rather than other noise rejection.

I'm not up to speed on what 'other noise' might be here. Yes seems Nige's amp was susceptible to the CM noise from the computer/monitor which was strongly blocked by the trafos.

I was thinking that noise other than CM noise transmitted on the signal lines might cause degradation in sound & the trafo may have rejected this due to it's bandwidth limitations? There may be LF or HF noise outside of the trafos pass bandwidth that would be rejected. Is this wrong thinking? Agreed the most likely culprit is CM noise but still worth bearing in mind other possibilities

Have you got a playback setup that emulates the SQ benefits heard in Nige's with the original trafo's (that you still have)?
This would enable some real world testing of the premise

I have an amp which doesn't already have trafos inside. But all my DACs have had trafos on the output from at least 3 years ago, so I suppose I could try with a smartphone or even with an el-cheapo SD-card player using a noisy PSU.

The CM chokes look far easier to wind being multi-segmented?

Easier than the trafos? Pretty similar, though here the wire's 0.1mm and could be even thicker.

The plan is to put the CM choke in series with the trafo to extend the isolation top end by a few octaves. I think I can reduce the trafo's capacitance by about an order of magnitude by means of the CM choke, so down to 10pF max. What I really need now is a supply of pre-twisted-pair wire so I can try out a bifilar CM choke....

Oh, I didn't realise this choke was meant to go in series with the trafo!